# tczhangzhi/pytorch-distributed

A quickstart and benchmark for pytorch distributed training.

Repository: https://github.com/tczhangzhi/pytorch-distributed
Canonical: https://ross.abutalabs.com/products/pytorch-distributed
Language: Python
License: MIT
License Family: permissive
Last push: 2024-07-25T10:57:27+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2449, "days_push": 769, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1655, forks 293 (observed 2026-08-28T04:05:17.809015+00:00)

## What it is
A collection of PyTorch example scripts demonstrating different distributed/multi-GPU training approaches (DataParallel, torch.distributed, torch.multiprocessing, Apex, Horovod, Slurm) on ImageNet. It also includes benchmark results comparing the speed of these methods on Tesla V100 GPUs.

## Use cases
- learn how to do multi-gpu training in pytorch
- compare pytorch distributed training frameworks
- quickstart code for torch.distributed
- benchmark apex vs horovod vs dataparallel
- set up distributed training on a slurm cluster
- speed up model training with multiple gpus

## When to choose
- you want copy-paste quickstart examples for PyTorch distributed training
- you need to decide which multi-GPU approach to use based on benchmarks
- you are learning distributed training concepts in PyTorch

## When to avoid
- you need a production-ready distributed training framework rather than examples
- you need up-to-date examples for the latest PyTorch APIs
- you need single-GPU training or CPU-only training

## Facets
- artifact type: learning-resource
- maturity: maintenance
- function: machine-learning, llm-training, benchmarking, developer-tools
- domain: deep-learning, machine-learning, tutorials, gpu-computing
- platform: python, cross-platform
- tags: pytorch, distributed-training, quickstart, horovod, apex, imagenet, multi-gpu, example-code, gpu, linux

## Member repositories
- tczhangzhi/pytorch-distributed (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:17.809015+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:44:53.685464+00:00, confidence not recorded.
  - readme: https://github.com/tczhangzhi/pytorch-distributed (fetched 2026-08-28T04:05:17.809015+00:00, sha b229a5c1187c)
- Data as of 2026-08-30T08:39:29.467469+00:00.